**** BEGIN LOGGING AT Fri Dec 20 03:00:11 2019 Dec 20 03:32:40 New news from stackoverflow: How to get the ARM tool chain with poky? Dec 20 07:15:24 I have come across a weird issue. I have a binary package that I want to install into my image. It is a tar.gz file which contains a whole bunch of binaries, libs, config files etc. All I really want is to extract it into a subdirectory under /. I created a recipe for this package, and it seems to accurately install all files under my desired directory. The problem is that when an application in this Dec 20 07:15:30 package is started, it hangs (the application is dotnet). But if I do not install the archive via the recipe, and instead just unpack the tarball on the target (flashed with the same image recipe without the binary package recipe) everything is fine. I compared the generated file systems, and found out that in addition to the subdir I want to install the binary package in, it also pulls in A LOT of other Dec 20 07:15:36 things all across the filesystem. Dec 20 07:16:09 The binary package has a few REPENDS to make QA happy, but even with these packages added to my "base" image, the diffenrence persists. Dec 20 12:24:14 kanavin: does AUH randomly pick a machine to do upgrades with? Dec 20 12:28:22 rburton, host machine or target machine? Dec 20 12:28:32 MACHINE Dec 20 12:28:51 AUH ran for acpica and then failed as acpica isn't compatible with qemumips :) Dec 20 12:28:57 it iterates over x86, x86_64, arm, mips, ppc and x86_musl for each upgrade Dec 20 12:29:06 yeah, so it's not a failure Dec 20 12:29:31 AUH could be taught to special case that maybe Dec 20 12:32:34 or just look at compatible_host etc Dec 20 12:34:18 i'll file a bug for reference Dec 20 13:13:33 kanavin: ok why can't i find auh in bugzilla? :/ Dec 20 13:13:40 ah found it Dec 20 13:16:56 rburton, I think my problem yesterday was recipe related, since OpenSSL crate in RUST picks the recipe-sysroot-native instead of recipe-sysroot as openssl lib path Dec 20 13:17:14 the path to openssl can be specified by env variable OPENSSL_DIR Dec 20 13:17:34 STAGING_EXECPREFIXDIR has the full path right? Dec 20 13:18:01 yes Dec 20 13:19:11 cool Dec 20 13:19:21 cmake seems to still want to find the openssl dir Dec 20 13:19:24 damn Dec 20 13:19:31 rburton, how would AUH determine not to run an update if host is incomaptible? is there something easy in bitbake -e that it can look for? Dec 20 13:19:58 e.g. MACHINE=not_compatible_machine bitbake -e acpica ----> what to look for? Dec 20 13:31:20 nvm, found out that won't work anyway Dec 20 13:51:24 rburton, LOL there are both recipe and cmakelists problems LOL :D 3 different openssl root path variables need to be specified for Dec 20 14:18:10 Ad0: awesome Dec 20 14:18:19 it compiles now :) Dec 20 14:18:26 one glorious day, openssl will ship a .pc file and everyone wil just use that Dec 20 14:18:28 thanks for pointing out the sysroot vs sysroot-native Dec 20 14:18:36 hehe Dec 20 14:25:29 rburton, also, they will manage API breaks like adults **** BEGIN LOGGING AT Fri Dec 20 17:25:31 2019 Dec 20 17:25:39 zeddii: using glibc from openembedded-core master to reproduce the problem? Dec 20 18:00:42 probably nothing to do with it but I did notice this: https://sourceware.org/git/?p=glibc.git;a=commitdiff_plain;h=eb502f72cda31fc9ebe37689491a0fdac93f67c8 Dec 20 18:10:39 georgem. I hadn't seen that, but I've only been looking at the kernel so far. Dec 20 20:08:25 hm I add DISTRO_FEATURES_remove = " alsa x11" but the bitbake -e myimage | grep ^DISTRO_FEATURES gives DISTRO_FEATURES_DEFAULT="acl alsa argp bluetooth ext2 ipv4 ipv6 irda largefile pcmcia usbgadget usbhost wifi xattr nfs zeroconf pci 3g nfc x11" Dec 20 20:09:04 why does bitbake even bother compiling the stuff I want to remove+ Dec 20 20:34:17 georgem. I just got mips64 to boot by hacking the gettimeof day routines. Dec 20 20:34:31 qemumips64 login: root Dec 20 20:34:31 root@qemumips64:~# uname -a Dec 20 20:34:31 Linux qemumips64 5.4.4-yoctodev-standard #1 SMP PREEMPT Wed Dec 18 16:14:02 UTC 2019 mips64 mips64 mips64 GNU/Linux Dec 20 20:34:39 now, I have to see if I can slean it up. Dec 20 20:34:44 s/slean/clean/ Dec 20 20:35:01 RP ^^^^^^^^^ Dec 20 20:35:38 zeddii: yay! Dec 20 20:35:49 zeddii: getting it to boot will be a relief! Dec 20 20:36:07 I basically forced it to stop reading a h/w clock and use a syscall. Dec 20 20:36:25 now, I can do a temp patch, report it to the list and move on with introducing the kernel as the replacements in master. Dec 20 20:36:41 * zeddii disliked learning anything about VDSO and mips clocks. Dec 20 20:37:17 I didn't want to go into Christmas holidays still with no idea on it. so it's great news for me at least. Dec 20 20:42:07 Congatec's gitlab doesn't allow anonymous cloning, and you can't sign up for an account. Not off to a good start is it? Can someone deduce from this repo which Linux kernel they're compiling in the reference build? https://git.congatec.com I'm not yet familiar with Yocto and without local checkout I can't grep Dec 20 20:45:37 oops sorry this is the correct URL https://git.congatec.com/yocto Dec 20 20:48:17 It is -very- old Dec 20 20:48:26 as in so old it's probably not useful Dec 20 20:49:00 kernel is 3.19 and it's before Sumo release.. thus pretty much useless Dec 20 20:49:26 ahh fido there we go.. that is the default Dec 20 20:50:38 fray, which file showed you this? Dec 20 20:51:08 poky/meta/conf/layer.conf shows no defined compatible string which started in sumo.. Dec 20 20:51:19 kernel poky/meta/recipes-kernel/linux/... Dec 20 20:51:29 and if yoiu look at the top of the screen it says the branch is named 'fido' ;) Dec 20 20:52:30 other repopsitories may be newer or old.. I was only looking in poky Dec 20 20:52:47 thank you Dec 20 20:55:20 so newest is Sumo branch with recipes for 4.12, 4.13 and 4.15. 4.15 is not too ancient, just barely meets my application's minimum requirements (4.14+) it's more the fact that they stopped there that scares me Dec 20 20:55:45 not exactly filling me with confidence to build on the next 8 years from this Dec 20 21:01:58 Is there a way for me in a recipe A to get the path to the build directory of recipe B? In my specific case, I have a linux kernel modules backports recipe and I need to pass KLIB_BUILD= to the make invocation. Dec 20 21:03:03 dfrey: usually no, but for kernel modules you can get such info via inherit module Dec 20 21:04:25 khem: Thanks. I'll take a look at what module.bbclass provides **** ENDING LOGGING AT Sat Dec 21 02:59:58 2019